Fix WSL systemd detection

[Syquel]

Does this PR introduce a user-facing change?

Fixes the detection of systemd in WSL.
Users who use WSL with systemd must recreate their machine to apply this fix.

If WSL is configured to use systemd by setting boot.systemd=truein /etc/wsl.conf, WSL creates a Symlink from /sbin/init to /usr/lib/systemd/systemd and starts systemd via /sbin/init.
podman machine start, podman machine ssh and other commands check if systemd is running via isSystemdRunning() and fail if it is not detected.

Because isSystemdRunning() incorrectly checks for the command name "/lib/systemd/systemd" instead of the executable path it does not detect systemd running as "/sbin/init".

[Luap99]

@n1hility PTAL

[n1hility]

Thanks for the PR @Syquel !

Is there a use-case you have for wanting to enable the built-in systemd? To support this properly a different bootstrap/start of machine would be required. The reason we don't yet have built-in support for this mode is because of a WSL bug that breaks user systemd services (used by podman), which is only fixed in preview releases (1.3.14+)

[Syquel]

Hi! I have been using WSL with systemd and podman for ~9 months now successfully (even before the official support) and after applying this patch the podman windows client works too.

My use case is the cgroupv2 support provided by systemd.

You are right that simply bootstrapping the WSL machine via podman machine init is not enough at the moment, but with a few minor customizations it works without any issues.
In my opinion podman machine init does way too much at the moment and it's way easier to deploy a "raw" fedora image to WSL and customize it to get postman to run properly with systemd, but that's an unrelated issue.

This patch simply enables the podman windows client to acknowledge that the podman "server" is running on a WSL machine with systemd enabled.
The current commands to determine if systemd is running are incorrect, because they look at the executable name in argv[0] instead of looking at the actually invoked executable.

As stated in my last review comment I am not sure about only checking PID 1 for systemd, but the general approach to check the actually invoked executable should be correct.

// EDIT
Currently I use the podman.exe generate by the build of this PR and I can finally use podman machine start, podman machine ssh etc again without any issues.

[Syquel]

Could someone take a look, please?

[baude]

code looks fine, what say you @n1hility

[TomSweeneyRedHat]

LGTM
but would like a head nod from @n1hility

[n1hility]

/lgtm

The change itself is fine so will merge it but until we have formal native systemd support can't promise there won't be other issues (don't consider it supported).

[rhatdan]

/approve

[openshift-ci]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: rhatdan, Syquel

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [rhatdan]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment